U.S. Patent No. 9,616,329: Adaptive learning system for video game enhancement
Issued April 11, 2017, to Electronic Arts, Inc.
Priority Date June 28, 2012
U.S. Patent No. 9,616,329 (the ‘329 Patent) describes a system for adaptive learning in a video game engine. AI opponents often fail to replicate the adaptive learning abilities of a human opponent. A human can quickly analyze an opponent’s strategy and adjust easily. The ‘329 Patent hopes to change that by providing a system to implement adaptive learning into a video game. According to the patent, the game will begin to record the gameplay of an initial match. Upon completion of the match, the system will analyze the recording for key moments and player habits which lead to the outcome. For example, in Madden, the game will analyze the sequence and moments that allowed the human player to score a touchdown. The resulting analysis and recorded sequence are stored within the system’s memory. During the second match, the system uses the recording to identify human player’s habits. Using the analysis from earlier, the computer opponent adjusts its tactics to counter the human player. The player may find that the running play that once scored touchdowns is no longer a viable option.
Techniques are disclosed that can capture sequences of player input in response to situations confronting a player of a video game while playing the video game on a game console. A game engine within the game console may associate an archetype with the captured sequences. The outcomes of the captured sequences are determined with respect to the game situations and a score is applied to the captured sequences based on the outcomes. The captured sequences, the outcomes and the scores are stored in a runtime memory within the game console used to play the video game. Upon completion of the video game, the stored captured sequences, the outcomes and the scores may be downloaded and stored to an external memory associated with the game console. Subsequent video game sessions may load the captured sequences, the outcomes and the scores into the runtime memory to be used in the next session.
1. A method comprising: during runtime execution of a video game application on a computing device, the video game application comprising a virtual environment, capturing sequences of player input in response to one or more virtual events confronting a player controlled virtual agent within the virtual environment of the video game application; for individual captured sequences, determining an outcome of the captured sequence with respect to a virtual event of the one or more virtual events; applying a score to the captured sequence based, at least in part, on the outcome; storing the captured sequence, the outcome, and the score in a runtime memory within the computing device used to play the video game application; in response to a virtual event confronting a computer controlled virtual agent within the virtual environment, identifying one or more captured sequences based, at least in part, on a game status and computer controlled virtual agent status associated with the virtual event confronting the computer controlled virtual agent; selecting a captured sequence from the one or more captured sequences based, at least in part, on the scores and outcomes associated with the individual sequences; and executing the selected captured sequence by the computer controlled virtual agent in response to the virtual event confronting the computer controlled virtual agent within the virtual environment.